green_chair

One thing I ran into is that people in any career path love it and believe it is the only oppropriate career path for you. So if you are asking professors for career advice, it will be hard to have that discussion as you've found out. Most graduate students don't seem to realize how their skills translate into jobs, so I have also gotten the boring responses you've gotten. What I found help was speaking to people outside of academia and to a career counselor (probably the only person who didn't have a stake in where I end up). Perspective is important here.

Comment on job prospects from someone who was on the market this season: There are lots of options. If your training is broad enough, the academic job options are in Psych, Education, and other departments. Ther past two years there have been about 35 tenure track job postings. There seems to be many places that are funding new data-focused research initiatives (I saw 4 positions related to something like this this year). Add to academia the huge demand for people with a quantitative training outside of academia and it doesn't seem like there will be a job shortage any time soon. So job prospects = good and far better than any other psych field.

R is where a lot of the interesting research is happening in and out of academia. There is nothing better than publishing a paper and a package side by side. But there is huge demand for Data Scientists in the overall job market which often calls for Python (or other programs beyond analyst-focused programs). But then again many jobs have operational code written in an older language, like SAS. What I have found is that a specific language is less important than being really good at one langagure (your bread and butter) and an ability to learn new languages as needed. The first language you learn is the hardest, so pick one and get it down. Then pick up others as the situation requires. As a counter-example, it would be silly to only know a "hello world" program in 30 different langauges. More than anything, being able to "teach" stats is very important. If you are in a position where you are the most stats heavy person, you have to be able to communicate to your boss and your bosses boss why these advanced statistical techiniques are important for their given problem.

I hope I never say no to that much money ever again. But I did accept another job, slightly lower salary, better fit and culture. Life after PhD isn't looking so bad

If we acknowledge that the more recent years are better predictors than the farther away years, then we can do better than regression. We have to make several assumptions (like an original prior, that the years are i.i.d. normal, and that weekends, holidays, and leap years don't play in the decision). Each year's prior is the previous year's posterior, and we update our estimate and our uncertainty with new data. Then the box plots of the posterior distributions for each year are on the left, and our prior (our best guess based on the data) for 2012 is on the right. The 95% HPD credible interval for the 2012 prior is the 93.63 to 98.53 day of non leap year, or from April 3rd to April 9th. And the most likely day is April 6th. So that doesn't narrow it down too much, but thats what I get from the data.